Table III.

Logistic Regression Models Predicting Whether Respondent Engaged in Unprotected Sex with a Particular Partner

	Each of the Five Variable Blocks Examined Separately	Final Model
	OR (95% CI), p-val	OR (95% CI), p-val
Partner Characteristics
Older Partner
Partner has been told he is HIV+
Partner has injected drugs
Partner has had sex with a man
Relationship Characteristics
Average frequency of sex in a typical month		1.04 (.98, 1.10)
How long woman has known partner		1.05 (.95, 1.17)
Frequency of partner’s tangible support		1.20 (.90, 1.61)
Frequency of arguments with partner		1.30 (.95, 1.77)
Partner has been physically abusive (vs. no abuse)		5.87 (1.81,18.92), p=.003
Partner has been psychologically abusive but not   physically abusive (vs. no abuse)		1.64 (.77, 3.50)
Woman’s relationship commitment		1.11 (1.01, 1.21), p=.03
Woman’s ability to refuse unwanted sex		.95 (.81, 1.11)
Drug/Alcohol Use with Partner Prior/During Sex
Partner drank alcohol before/during sex with her
She drank before/during sex with her partner
Partner used drugs before/during sex with her
She used drugs before/during sex with her partner
HIV and Condom Attitudes
Positive attitudes towards condoms		.20 (.10, .40), p<.001
Higher HIV susceptibility		1.62(1.21, 2.17), p=.001
Social Network Characteristics
Percentage of non-sex partner alters perceived as  engaging in risky sex		.99 (.97, 1.00)
Percentage of non-sex partner alters with whom she   has discussed HIV/AIDS		.97 (.96, .99), p=.006
Network density		4.32 (.82, 22.61)

Note. All models control for women’s race/ethnicity, age, education, employment status, and whether pregnant or tried to become pregnant in past 6 months. Additional models were run to test for associations between type of recruitment site and the dependent variable and there were no significant changes to the findings.